Arm adds virtual testing platform for Corstone Cortex-A and -M ref designs

Arm adds virtual testing platform for Corstone Cortex-A and -M ref designs

Oct 18, 2021 — by Eric Brown

49 views

Arm has launched an “Arm Total Solutions for IoT” initiative that combines its Cortex-A- and -M based Corstone SoC reference designs with a new “Arm Virtual Hardware Targets” platform for virtual, cloud-based testing.

In 2018, Arm launched an Arm Corstone subsystem product line comprising prevalidated SoC reference designs that combine its core IP with security, debug, memory subsystems, and in some cases, NPUs. Now, Arm has expanded Arm Corstone with a cloud-based Arm Virtual Hardware Targets testing and development platform aimed at IoT. Together with a new Project Centauri ecosystem initiative for Cortex-M developers, the programs are wrapped up into an “Arm Total Solutions for IoT” umbrella platform.

Separately, Arm announced an Arm 5G Solutions Lab in partnership with Tech Mahindra to enable Arm’s hardware and software ecosystem partners to demonstrate end-to-end 5G solutions in a live test environment.

Arm Total Solutions for IoT conceptual diagram
(click image to enlarge)

So far, there have been more than 150 Arm Corstone designs from Arm silicon partners, with nearly 70 percent of Arm’s Cortex-M55 licensees using the subsystems, says Arm. Although most of these have been based on Cortex-M. there are two Corstone designs for Linux-driven Cortex-A SoCs: the Corstone-500 (Cortex-A5) and Corstone-700 (Cortex-A32) subsystems. Arm also revealed an upcoming Arm Corstone-1000 platform that will be supported by Arm Virtual Hardware Targets that combines Cortex-A53 and Cortex-M cores (see farther below).

Whereas Arm Corstone is aimed primarily at SoC hardware developers, especially smaller fabless design firms building smaller runs of custom SoCs, Arm Virtual Hardware Targets is intended primarily for the software developers, OEMs and service providers writing code for the SoCs.

“Imagine if every mobile app developer had to buy every phone model to test their software,” said Mohamed Awad, VP of IoT and Embedded at Arm at a recent press briefing. Arm Total Solutions for IoT aims to approximate the mobile app development experience, said Awad. This includes hardware and software developed in parallel, cloud based testing, and APIs to increase software reuse over diverse hardware targets, he added.

Arm Total Solutions for IoT proposes to reduce product design cycles by up to two years. “Through a radical change in how systems are designed, Arm is uniquely positioned to fuel a new IoT economy that rivals the shape, speed and size of the smartphone industry’s app economy,” stated Awad in today’s announcement.

Arm Total Solutions for IoT is part of a growing outreach by Arm to smaller chip designers and their ecosystem partners who are driving the market for custom SoCs for proliferating use cases enabled by wireless IoT and edge AI. Chip designers that have hitched their wagon to the open RISC-V platform are aiming for a foothold in these markets, as is Intel with its Intel Foundry Services for assisting semiconductor startups. Arm has previously addressed smaller SoC designers with its Arm Flexible Access licensing program and Custom Instructions extensions.

Arm Virtual Hardware Targets

The cloud-based Arm Virtual Hardware Targets platform offers a virtual model of the Corstone subsystem to enable software development without requiring a time-consuming investment in hardware farm testing. The platform “runs in a simple application in Linux environment” and provides agile software development methodologies like continuous integration / continuous deployment (CI/CD), DevOps, and MLOps, for IoT and embedded platforms, says Arm.

Arm Virtual Hardware Targets provides models of Arm-based SoCs with mechanisms for simulating memory, peripherals and more, to “reduce a typical product design cycle from an average of five years to as little as three years,” claims Arm. This virtual approach enables Arm silicon partners to integrate customer feedback into their designs before tape out, “while enabling the entire IoT value chain to easily develop and test code on the latest IP well ahead of silicon availability.”

Arm Virtual Hardware Targets workflow diagram (left) and roadmap
(click images to enlarge)

Available now is a Virtual Hardware Target for the Arm Corstone-300 subsystem, which combines an Arm Cortex-M55 with a 0.5-TOPS Arm Ethos-U55 microNPU. The virtual target, which “addresses general-purpose compute and ML workload use-cases, including an ML-based keyword recognition example,” is already being used by Arm partners.

The next round of Virtual Hardware Targets will support the upcoming, Linux-ready Cortex -A53/Cortex-M Corstone-1000 design. There will also be a keyword recognition focused target for the Cortex-M33 based Corstone-200 and a future voice recognition target for an upcoming Polaris Corstone with a new Olympus chip and an Ethos-U55 core. A third round will offer a vision-oriented spin of the Corstone-1000, as well as a smart sensor fusion target with a future Cortex-M based Corstone and an object recognition target for Olympus combined with an upcoming Ethos follow-on code-named Zaphod.

Arm Total Solutions for IoT also encompasses a new Project Centauri platform billed as a Cortex-M counterpoint to the Cortex-A focused Project Cassini. Project Centauri delivers a set of device and platform standards, as well as reference implementations for device boot, security, and cloud integration.

Arm Corstone for Linux-driven Cortex-A

Arm Corstone, which is supported by the Arm Flexible Access licensing program, currently comprises four Cortex-M reference designs and two Cortex-A designs. These will eventually be joined by the Cortex-A53 based Corstone-1000

The Corstone-500 (SSE-500) is built around the low-powered Cortex-A5 core and is available with or without system IP. The design appears to be a follow-on to Arm’s Cortex-A5 DesignStart program from 2018.

Block diagrams for Arm Corstone-500 (left) and Corstone-700
(click images to enlarge)

The full design includes an SRAM controller, AXI interconnect, CoreLink L2 cache, a CoreLink NIC-4000 network interconnect, and CoreSight SoC-400 debug and trace components. Peripheral support includes PrimeCell UART, synchronous serial, and GPIO interfaces, as well as an RTC, watchdog, and True Random Number Generator (TRNG), among other features.

The Corstone-700 (SSE-700) taps the rarely used Cortex-A32 along with up to two Cortex-M cores, with a default to Cortex-M33 via a Corstone-200 (SSE-200) subsystem. Cortex-A32 may have been skipped over by chip designers due to its hybrid 32-bit-only implementation of ARMv8. When it was announced in 2016, the up to quad-core Cortex-A32 was said to offer higher performance per mW than Cortex-A7.

The Corstone-700 provides a secure enclave with Root-of-Trust, and supports expansion with sensors, connectivity, video, audio, and machine learning technology. Other features include AXI, secure debug, power and system control, firewalls, RTC, TRNG, and more.

Further information

The first configuration of an Arm Total Solution for IoT is available now on AWS Marketplace is the form of Arm Virtual Hardware Target for the Corstone-300. More information may be found in Arm’s announcement and this Arm Virtual Hardware page.